In this paper, we develop a multi-level comparative approach to analyse Trends in International Mathematics and Science Survey (TIMSS) and Programme of International Student Achievement (PISA) mathematics results for a country, Russia, where the two tests provide contradictory information about students' relative performance. Russian students do relatively well on the TIMSS mathematics test but relatively poorly on the PISA. We compare the performance of Russian students with different levels of family academic resources over the past decade on these tests compared to students with similar family resources in Russia's neighbours and to Russian students studying in Latvian and Estonian Russian-medium schools. These comparisons and interviews with educators in Latvia and Estonia help us understand why students in Russia may perform lower on the PISA and to draw education policy lessons for improving international test performance generally and Russian students' PISA mathematics performance specifically.